RECAP-Forcing: Retaining Content Appearances for Long Video Generation

Hugging Face Daily Papers Papers
long-video-generation appearance-novelty memory-indexing kv-cache training-free video-consistency optical-flow

Summary

RECAP-Forcing improves long video generation by indexing memory based on appearance novelty instead of recency, preserving key-value caches for new content to maintain consistency and quality without additional training.

Long autoregressive video generation faces a fundamental memory challenge: with a finite attention window, a model must decide which information from an ever-expanding history to retain. Existing methods organize memory temporally, preserving recent frames while compressing or discarding older ones. We instead propose RECAP-Forcing, organizing memory by appearance novelty. A long video is not merely a sequence of frames, but an evolving cast of subjects, objects, and scenes whose identities must remain consistent over time. We organize memory by retaining the KV cache associated with newly appearing content--such as entering subjects, disoccluded regions, and newly introduced scenes--at the moment it first becomes visible, prioritizing novelty over recency. Memory should scale with the amount of newly introduced content, rather than with video length. This appearance-indexed memory makes long-range consistency an explicit property of the memory structure. Our framework unifies two mechanisms under this single principle. At the beginning of a video, when all visible content is novel, an attention sink preserves the initial scene. As the video evolves, an optical-flow-based novelty bank extends the same principle by selectively retaining newly revealed content. As a training-free inference method with no additional learnable parameters, RECAP-Forcing consistently improves visual quality and semantic fidelity across multiple strong baselines and outperforms existing memory methods.
Original Article
View Cached Full Text

Cached at: 09/01/26, 11:43 PM

Paper page - RECAP-Forcing: Retaining Content Appearances for Long Video Generation

Source: https://huggingface.co/papers/2608.26671

Abstract

RECAP-Forcing improves long video generation by indexing memory according to appearance novelty rather than recency, preserving key-value caches for newly visible content to maintain long-range consistency without extra training.

Longautoregressive video generationfaces a fundamental memory challenge: with a finiteattention window, a model must decide which information from an ever-expanding history to retain. Existing methods organize memory temporally, preserving recent frames while compressing or discarding older ones. We instead proposeRECAP-Forcing, organizing memory byappearance novelty. A long video is not merely a sequence of frames, but an evolving cast of subjects, objects, and scenes whose identities must remain consistent over time. We organize memory by retaining theKV cacheassociated with newly appearing content--such as entering subjects, disoccluded regions, and newly introduced scenes--at the moment it first becomes visible, prioritizing novelty over recency. Memory should scale with the amount of newly introduced content, rather than with video length. This appearance-indexed memory makeslong-range consistencyan explicit property of the memory structure. Our framework unifies two mechanisms under this single principle. At the beginning of a video, when all visible content is novel, anattention sinkpreserves the initial scene. As the video evolves, anoptical-flow-based novelty bankextends the same principle by selectively retaining newly revealed content. As a training-free inference method with no additional learnable parameters,RECAP-Forcingconsistently improves visual quality and semantic fidelity across multiple strong baselines and outperforms existing memory methods.

View arXiv pageView PDFProject pageGitHub8Add to collection

Get this paper in your agent:

hf papers read 2608\.26671

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2608.26671 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2608.26671 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.26671 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Echo-Forcing: A Scene Memory Framework for Interactive Long Video Generation

Hugging Face Daily Papers

Echo-Forcing introduces a scene memory framework for interactive long video generation, using hierarchical temporal memory, scene recall frames, and difference-aware memory decay to handle prompt switching and long-term recall. The method is training-free and achieves strong performance on VBench-Long.

KeyRec: Bounded Visual Memory for Streaming and Long-Video Understanding

Hugging Face Daily Papers

KeyRec is a training-free framework for building bounded visual memory in vision-language models, using a recent-observation cache plus a structured event bank to compress long/streaming video context to just 10% of dense visual tokens while setting state-of-the-art results on 13 of 15 streaming and long-video benchmark settings.

Memento: Reconstruct to Remember for Consistent Long Video Generation

Hugging Face Daily Papers

Memento is a subject-reconstruction-guided framework that improves long-form video generation by preserving recurring subjects through memory-based reconstruction and dual-query mechanisms, achieving state-of-the-art performance in long-term subject consistency and cross-shot coherence.